I use your train.prototxt to train face landmark with caffe. the loss converge to landmark mean value and not change. how to solve this problem. thx. - Githubissues

guozhongluo / head-pose-estimation-and-face-landmark

head pose estimation

248 stars 99 forks source link

I use your train.prototxt to train face landmark with caffe. the loss converge to landmark mean value and not change. how to solve this problem. thx. #14

Open nan10921p opened 6 years ago

jiangziya commented 6 years ago

I have the same problem.